Competitive intelligence for Virginia commercial real estate — before it hits the trade press.
PropIntel is an automated intelligence platform that tracks permit filings, team changes, and development announcements across Virginia's CRE market, and surfaces actionable signals in one place.
Most CRE professionals learn about competitor moves after they're public. PropIntel flips that. It continuously crawls public data sources — permit portals, company websites, careers pages, and press releases — resolves messy LLC names back to their real developer parent, and delivers clean signals to your inbox or Slack before anyone else picks up the story.
| 🏢 20+ firms tracked | 📄 419+ permits ingested |
| ⚡ 148+ signals generated | ✅ 0 misattributions |
Crawls permit portals, team pages, careers pages, and press releases with Playwright-backed fallback for JavaScript-heavy sites. A 6-rung entity resolution system maps raw permit applicant LLCs to known parent developers — zero misattributions across 395 name mentions.
Ask questions about any tracked firm and get grounded answers with source URL and page-level citations. Powered by 384-dimension embeddings (ONNX Runtime) and pgvector HNSW indexing under the hood.
Four automated signal detectors run continuously:
- Hiring surge — 1.5× above baseline
- Asset-class pivot — 30%/10% shift threshold
- Geographic expansion — first permit in a new locality
- Permit volume anomaly — Z-score statistical spike
Get notified via email or Slack the moment a signal fires. Every week, an AI-written digest summarizes each firm's activity in plain English — no dashboards to check.
Hiring moves, permit filings, press announcements, and derived trend signals all land in a single feed. Alerts, digests, and firm timelines read from the same source — no duplicate data, no drift.
Per-tier JWT + API key authentication with Redis-backed atomic rate limiting. Built for multi-tenant use from day one.
Public Data Sources
├── Fairfax County ArcGIS
├── Socrata Open Data
├── Company Websites
└── Trade Press
│
▼
Ingest Layer
├── Depth-Aware Crawler
├── Playwright Fallback
├── Bot-Challenge Detection
└── PDF Extraction
│
▼
Intelligence Engine
├── Entity Resolution (6 rungs)
├── Embedding Pipeline (384-dim)
├── Trend Detection (4 detectors)
└── RAG Q&A Service
│
▼
Output Surfaces
├── Firm Timelines
├── Semantic Search
├── Weekly AI Digest
└── Alert Delivery (Email + Slack)
| Layer | Technology |
|---|---|
| Backend API | FastAPI + Pydantic, Python 3.11+ |
| Database | PostgreSQL 17 + pgvector 0.8.6 (Neon) |
| Cache / Rate Limiting | Redis |
| Migrations | Alembic (HNSW + trigram indexes) |
| Frontend | React + TypeScript (Vite) |
| AI / Embeddings | fastembed (ONNX Runtime), Claude API |
| Notifications | Resend (email), Slack |
| Crawling | Custom WebScraper core + Playwright |
Curious about the build, want a walkthrough, or thinking about something similar for your market?
Areeb Arshad · areebarshad@vt.edu
Happy to chat about the architecture, the data pipeline, or anything in between.
Built for Virginia CRE — open source under MIT.